technical white paper certification of eternus dx in ... · pdf filetechnical white paper...

White paper

of 14 http:// www.fujitsu.com/eternus/

Technical White paper Certification of ETERNUS DX in Milestone video surveillance environment

This white paper describes the certification tasks for Milestone video surveillance solutions with ETERNUS DX storage systems

White paper [Certification of Milestone XProtect Corporate software solutions for video surveillance with ETERNUS storage systems]

of 14 http://www.fujitsu.com/eternus/

Table of Contents

1. Introduction 3

2. Configuration 3

2.1 Recording Server 4

2.2 Management Server 4

2.3 ETERNUS DX200 S3 4

2.4 Smart Clients 6

3. Test Types 6

3.1 Benchmark test 6

3.2 Maximum performance test 9

4. Calculation for DX200 S3 max configuration 12

5. Calculation for DX600 S3 max configuration 13

6. Conclusion 14



1. Introduction Milestone XProtect Corporate is the ultimate software for video surveillance in large-scale and high-security areas, such as airports and city surveillance. XProtect Corporate is powerful IP video management software (VMS) designed for large-scale and high-security deployments. Its single management interface enables the efficient administration of the system including all cameras and security devices regardless of its size or if it is distributed across multiple sites. For systems demanding supreme situational awareness and precise response to incidents, XProtect Corporate features interactive maps linked to alarms and included support for XProtect® Smart Wall. XProtect Corporate provides the ultimate system reliability for high-security installations. Edge Storage support combined with failover recordings and redundant management servers ensure video recordings are never interrupted. This white paper describes the required tasks to have ETERNUS storage systems certified for Milestone environment.

2. Configuration Fujitsu provided Milestone technical information about ETERNUS DX200 S3 and ETERNUS DX600 S3 and got a recommendation for Storage and Server configuration. Five recording server, one management server and 180 disks within ETERNUS DX200 S3 are planned for the configuration. Additional five virtual machines for Smart Clients are used to view the cameras stream from the Live-DB and Archive-DB. The Smart Clients use a dedicated network (orange lines, 10Gb/s) for viewing the stored video streams.



2.1 Recording Server The recording servers are based on four PRIMERGY RX2540M1 and one PRIMERGY RX300S8 with the following configuration: PRIMERGY RX2540 M1

2 x Intel® Xeon® CPU E5-2620 v3 @2.40Ghz

128 GB MEM

Emulex LPe16002

Windows Server 2012 R2 PRIMERGY RX300 S8

2 x Intel® Xeon® CPU E5-2630L v2 @2.40Ghz

64 GB MEM

Emulex LPe16002

Windows Server 2012 R2 The recording servers are connected with 2 ports of 16Gb/s Emulex fibre channel HBAs to the Brocade switch. A “crossover” zoning is done on the switch, so each HBA ports sees 4 ports on the ETERNUS storage.

2.2 Management Server The management server is based on one PRIMERGY RX4770 M1 with the following configuration: PRIMERGY RX4770 M1

4 x Intel® Xeon® CPU E&-4880 v2 @2.50Ghz

128 GB MEM

2.3 ETERNUS DX200 S3 The ETERNUS DX200 S3 is connected with four 16Gb/s fibre channel ports to the ETERNUS switch. The raid configuration is done as follow: Live-DB

2 x Raid10 Raid Group (16+16 disks) total 64 disks (300 GB SAS, 10Krpm)

Vol00 and Vol01 are created in each RAID10 raid group and provided to Recording Server RS-01







The two volumes are “software striped” within each recording server. The benefit of the software stripe is that access to the volume is done via the preferred path. This guaranteed a proper I/O load balancing and equal distributed cache memory on each controller module (CM). The internal stripe size for Raid10 is 64KB.

Archive-DB

4 x Raid6-FR Raid group ((12D+2P)x2+1HS disks) total 116 disks (1 TB NL-SAS, 7200rpm)

Vol02 - Vol05 are created in each RAID6-FR raid group and provided to Recording Server RS-01




Vol42 - Vol45 are created in each RAID6-FR raid group and provided to Recording Server RS-05 The four volumes from the Raid6-FR Raid group are “software striped” within each recording server. The benefit of the software stripe is that access to the volume is done via the preferred path. This guaranteed a proper I/O load balancing and equal distributed cache memory on each controller module (CM). Raid6-FR is a special raid group in ETERNUS DX storage systems and is used to have a fast rebuild in case of a disk failure within Raid6. The internal stripe size for Raid6-FR is 64KB. The picture below shows the disk configuration.



2.4 Smart Clients Additional virtual machines are created on a Primergy server RX4770 M1 with RHEL 7.1 operating system. The virtual machines are based on KVM hypervisor and are used for Smart Clients. With Smart Clients you can view the live recording data or playback data from the past.

3. Test Types

The load tests which are included in the basic scope of the certification tests includes

Benchmark test

Maximum performance tests

3.1 Benchmark test

The benchmark tests are performed with 100 cameras on each recording server with the following parameter:

Each simulated camera created a data stream of about 780 KB/s. For 100 cameras on each recording server 78 MB/s are created. The picture below shows an average bandwidth of 773.504 bytes for one camera.



When activating 100 cameras on each recording server, the ETERNUS performance monitoring tools PMCC shows that each recording server generates roughly 76 MB/s (2 x 38 MB/s) write I/O. The response time is about 7-8 ms for write I/O.

After establishing the write I/O to the Live-DB, 25 cameras were “Live viewed” on each smart client. This will forced additional LAN throughput of about 20 MB/s for each smart client. Below you can see the LAN performance for one smart client after starting the live view.



With starting of the “playback function”, 25 videos are read from the Live-DB on each recording server. This results in a read throughput of about 16 MB/s (2 x 8 MB/s) per recording server. The read latency is about 3-4 ms.

On the right side you can see the disk busy rate for this benchmark test. In total 180 disks are involved, so you will only see an extract of all the disks. The disks with a busy rate of about 70% belong to the Live-DB. Result: The result of the benchmark test is that write I/O of 500 cameras and playback read I/O from 100 cameras will cause an overall throughput of 450 MB/s. The throughput on the LAN for communication with the smart clients is about 100 MB/s. The disk busy rate is about 70 %, so much more bandwidth is available from ETERNUS DX point of view.



3.2 Maximum performance test

The maximum performance tests are used to prove the overall solution, including the Archive-DB. The settings are done as follow:

75 cameras per recording server

Check every full hour if data is older than 1 hour in the Live-DB

Archive recordings older than 1 hour at the next archive schedule

Use the following stream settings



The test schedule is done as follow:

- 10:53 Start tests - 11:00 Start performance monitoring (perfmon) and PMCC - 12:00 Check if data is older than 1 hour in Live-DB move data from 10:53 – 11:00 into Archive-DB - 13:00 Check if data is older than 1 hour in Live-DB move data from 11:00 – 12:00 into Archive-DB

Confirm if estimated copy process takes no longer than 1 hour - 14:00 Check if data is older than 1 hour in Live-DB move data from 12:00 – 13:00 into Archive-DB - 14:10 Start Playback function, view Videos from Archive (11:00 – 12:00) - 15:00 End test

Below you can see an example of an archiving task. The copy process from Live-DB into Archive-DB started at 1:00 pm and the estimated end time is 1:42 pm. This means all data older than 1 hour is copied within 1 hour into Archive-DB. It is important that the estimated time in no longer than 1 hour. Otherwise the Live-DB will be overloaded and video streams may be lost. Notice: The critical limit when copying data from Live-DB into Archive-DB is reached with 80 cameras. This time takes about 60 minutes, so it may happened that not all data are saved into Archive-DB. This limit is not related to storage or server bandwidth, but it is more related to the XProtect software pipeline. The software needs to analyze and provide metadata (time, device ID) for each frame before it is sent to disk.

At 14:10 additional playback streams are started. This means that videos stored in the archive are viewed. This process causes additional I/O traffic from the Archive-DB. Below is the confirmation that the backup is still possible within one hour (estimated end time = 2:40 PM).



The PMCC data shows that the read throughput from the Live-DB (Vol00 – Vol01) for storing the data into Archive-DB (Vol02-Vol05) is higher than the write throughput created from the cameras. The read throughput from the Archive-DB is caused from the playback function. This is about 20 MB/s for 25 cameras.



Result: The overall result of the maximum performance test is as follow:

375 cameras write to Live-DB (10 x 26 MB/s): ~ 260 MB/s

Archive process read from Live-DB (10 x 35 MB/s) and write into Archive-DB (20 x 18 MB/s): ~ 710 MB/s

Read from Archive-DB via playback function (20 x 4Mb/s) ~ 80 MB/s

---------------- ~ 1050 MB/s ========== The maximum performance tests created a total I/O bandwidth of about 1050 MB/s. The throughput on the LAN is nearly 100 MB/s. On the right side you can see the disk busy rate of about 60 %, so much more bandwidth is available from ETERNUS point of view. Of course the I/O throughput is one important indication of the overall configuration, but the “frame loss” indication is more important. The Windows perfmon logs showed that the frame loss rate is in a really good shape, 0.11% or less for the benchmark tests and 0.06 % for the maximum performance tests. Milestone confirmed that these values are really good.

4. Calculation for DX200 S3 max configuration Assumption:

1) In the normal operation phase the overall performance is 1050 MB/s

2) Disk busy rate is about 60%, but let’s calculate with 70% to have more confidence. Disk busy factor (DBF) = 100% / 70 % = 1,42

3) 180 disks are used, 264 disks are possible, disk factor (DFA) 1,46 (264/180)

4) Cameras actual in use = 5 x 75 = 375 (CAM)

Formula for maximum cameras: DBF x DFA x CAM = Max CAM 1,42 x 1,46 x 375 = 777

Result: DX200 S3 roughly may support 777 cameras.



5. Calculation for DX600 S3 max configuration The tests for ETERNUS DX600 S3 are not performed because Milestone recommends more than 1000 disks for certification tests and at least 7 recording server. Based on the test result from ETERNUS DX200 S3 and the fact that DX600 S3 can have 4 times more disks installed (1056) and distributed over 4 backend loops, the following estimated performance may be reached with DX600 S3.

Assumption:

1) In the normal operation phase the overall performance is 1050 MB/s

2) Disk busy rate is about 60%, but let’s calculate with 70% to have more confidence. Disk busy factor (DBF) = 100% / 70 % = 1,42

3) 180 disks are used, 1056 disks are possible, disk factor (DFA) 5,86 (1056/180)

4) Cameras actual in use = 5 x 75 = 375 (CAM)

Formula for maximum cameras: DBF x DFA x CAM = Max CAM 1,42 x 5,86 x 375 = 3120

Result: DX600 S3 roughly may support 3120 cameras



6. Conclusion ETERNUS DX S3 storage systems completely fulfill the requirements to work in the Milestone XProtect Corporate environment for video surveillance on large-scale and high-security areas. The Live-DB and the Archive-DB can be coexisting within one ETERNUS DX system. A total capacity of 1584 TB (DX200 S3) or 6336 TB (DX600 S3) can be considered for planning the configuration. Please take into account that the throughput of Live-DB is higher compared with the throughput of Archive-DB. It is recommended that the Live-DB should be designed with SAS disk drives and the Archive-DB can be designed with NL-SAS disk drives.

Contact FUJITSU Limited Website: http://www.fujitsu.com/eternus/

Not all System names or product names in this document are followed by (R) or (TM). Other company, product names may be trademarks or registered trademarks of their respective owners. This white paper is devoted to provide technical information and an overview of the basic facilities of ETERNUS Disk Storage Systems in Milestone XProtect Corporate environment. The contents of this document may be modified without any prior notice. Please contact FUJITSU Limited if you find any error in description. FUJITSU Limited is not responsible indemnity that might be caused by the contents in this documentation or any damage related to contents in this documentation. Copyright 2016 FUJITSU LIMITED

http://www.fujitsu.com/eternus/

technical white paper certification of eternus dx in ... · pdf filetechnical white paper...

Documents